Post Editing – Getting Out What You Put In

As a communications professional I’m always impressed by the volumes of live information available and how it’s all compiled in the first place. In fact, any marketing professional who claims not to have used Wikipedia as a general information source in recent years, when writing press releases or articles would, in my opinion, be hard to find.

This is why I was particularly intrigued to read an article about launching its Health Speaks crowd-sourcing initiative recently, which calls upon volunteers to edit machine translations of health information found on sites such as Wikipedia into Arabic, Hindi and Swahili.

Crowd sourcing is obviously becoming more popular across many industries to gain opinion, share knowledge and collate information, and it was only a matter of time before translation would actively come into the mix.  But audiences need to bear in mind that the information collated in Wikipedia, the source from which the Google Translator Toolkit and volunteer linguists will provide the translations, won’t always be factually accurate.

While I applaud this method of generating content from community contributions, there is a particular risk in relation to the translation of health information.  Incorrect medical translations could lead to embarrassing mistakes.  Worst case scenario: a poor translation could potentially become a health risk.

Machine translation has gained real traction in recent months with major organisations putting their trust and their budgets into a post edited version, where the translation output is amended in terms of spelling and grammar and glossaries are applied to ensure it makes sense to the intended audience.

Any professional linguist, which I assume the volunteer editors will be, is trained to rely on the accuracy of the source material.  As a Chinese translation company, we spend a lot of time ensuring that we on-board customers properly and this includes checking the quality of the source material and simplifying it where necessary in order for the translations to make more sense.  We also insist on having target language glossaries in place to ensure technical terms are accurately understood and therefore translated by the linguist.  I would be interested to know how much source checking will be done by Health Speaks as the initiative builds momentum and more languages (and more detailed health material) are introduced to the initiative.

It looks as though Google is sticking to the right kind of content to avoid any of these problems for now – with a focus on health tips, disease prevention and dietary advice, rather than symptom-driven diagnosis.  We can expect more organisations to jump on the crowd sourcing bandwagon and use machine translations rather than human linguists, which could present a quality issue that may need to be addressed further down the line.  When translating medical information it’s critical that the translations are accurate and, as we say time and time again, a translation is only as good as the source material. Google has done the right thing and only selected pages that have been reviewed, but is it being verified by a medical professional?

Risks (which I am confident Google has taken into account) aside, this is a really interesting project with a genuine opportunity to help local charities through donation incentives.  So, as long as the audiences remembers that the information published is not necessarily from a qualified medical practitioner then the initiative will provide a good starting point for people looking for health advice.

So, let’s watch this space for now and commend Google on its contribution.